Weight loss through bariatric surgery is efficient for treatment or prevention of obesity related diseases such as type 2 diabetes and cardiovascular disease. Long term weight loss response does, however, vary among patients undergoing surgery. Thus, it is difficult to identify predictive markers while most obese individuals have one or more comorbidities. To overcome such challenges, an in-depth multiple omics analyses including fasting peripheral plasma metabolome, fecal metagenome as well as liver, jejunum, and adipose tissue transcriptome were performed for 106 individuals undergoing bariatric surgery. Machine leaning was applied to explore the metabolic differences in individuals and evaluate if metabolism-based patients’ stratification is related to their weight loss responses to bariatric surgery. Using Self-Organizing Maps (SOMs) to analyze the plasma metabolome, we identified five distinct metabotypes, which were differentially enriched for KEGG pathways related to immune functions, fatty acid metabolism, protein-signaling, and obesity pathogenesis. The gut metagenome of the most heavily medicated metabotypes, treated simultaneously for multiple cardiometabolic comorbidities, was significantly enriched in Prevotella and Lactobacillus species. This unbiased stratification into SOM-defined metabotypes identified signatures for each metabolic phenotype and we found that the different metabotypes respond differently to bariatric surgery in terms of weight loss after 12 months. An integrative framework that utilizes SOMs and omics integration was developed for stratifying a heterogeneous bariatric surgery cohort. The multiple omics datasets described in this study reveal that the metabotypes are characterized by a concrete metabolic status and different responses in weight loss and adipose tissue reduction over time. Our study thus opens a path to enable patient stratification and hereby allow for improved clinical treatments.